Turning implicit into explicit: does it matter if the rainbow color schemes are continuous or discrete? A case study with 2D and 3D permafrost visualizations

We report on a user study (<em>n</em> = 184) evaluating how discretization might affect reading values from permafrost visualizations with the rainbow color scheme (RC), and how it might interact with selected visualization and task types. After collecting data from 20 experts on their preferences, we compare continuous vs. discrete (banded) versions of the popular yet controversial rainbow colors. As visualizations, we include <em>2D</em> vs. <em>3D</em> versions, and a condition in which <em>context</em> is represented. We examine two kinds of lookup tasks: involving a single value, and involving a series of values. The outcomes of the study show a performance improvement with the discrete (banded) color scheme compared to continuous one on all objective and subjective metrics: response accuracy, response time and user satisfaction for all tasks except one. Discretization also moderates participants' performance: they perform better with <em>2D</em> with continuous colors, but their challenge reading in <em>3D</em> diminishes with discrete colors. Elementary lookup tasks appear to be harder on average than synoptic lookup tasks. Our results suggest that strictly following the <em>expressiveness principle</em> is not always advisable. Also importantly, they demonstrate that careful consideration of interactions between various factors is necessary in assessing user performance for reading and analyzing (geo)visualizations.